Direct Knowledge and Other Minds
Theoria, vol. LXX, part 2-3, pp. 271-293
Åsa Maria Wikforss
Stockholm University
[email protected]
1
Direct Knowledge and Other Minds
Åsa Maria Wikforss
The notion of direct knowledge is central to John McDowell’s philosophy. It is of importance
not just for his account of empirical knowledge but also for his account of rule-following,
singular thought, the past, ethics, and other minds. The reason McDowell appeals to the
notion of direct knowledge is that he wishes to oppose certain inferential models that he
thinks are pernicious and lead to skepticism.1 For instance, in Mind and World he argues that
unless we grant that there is a direct point of contact between our empirical judgments and
the world, we will end up with a holism that threatens to undermine not only empirical
knowledge, but also the very idea of empirical content.2
McDowell’s reliance on the notion of direct knowledge poses a certain challenge. The
problem is to present an account of direct knowledge which is both epistemologically
significant (for example, it should allow for an interesting distinction between knowledge that
is direct and knowledge which is not) and, at the same time, makes it true that the kinds of
knowledge he wants to construe as direct come out as such. The purpose of this paper is to
determine whether McDowell can meet this challenge in the case of other minds. Is there a
construal of direct knowledge which is both non-trivial and yet makes it true that knowledge
of other minds is direct? The claim that knowledge of the external world is direct is relatively
familiar, but the suggestion that we can know the mental states of others directly is less so
and few people have been convinced by it.3 Knowledge of other minds seems to be a prime
example of inferential knowledge. However, McDowell argues that construing knowledge of
other minds as indirect, inferential, leads to skepticism and fails to give a plausible account of
our use of mental concepts. We must, he suggests, make a radical break with the traditional
picture of the mind as ‘inner’ and hidden behind the ‘outer’, mere behaviour, and this
requires saying that the fact that another person is in a certain mental state is open to direct
1
It should be noted that McDowell’s goal is not so much to give a reply to the skeptic as to suggest a picture
which makes the skeptical challenge lose its urgency. In Mind and World he says: “The aim here is not to
answer skeptical questions, but to begin to see how it might be intellectually respectable to ignore them.”(1994,
113)
2
McDowell 1994.
3
For a critical discussion of McDowell on other minds see for instance Bilgrami 1992, Peacocke 1984, and
Robinson 1991.
2
observation: “On a suitable occasion, the circumstance that someone else is in some ‘inner’
state can itself be an object of one’s experience.”4
My question, therefore, is whether McDowell is right to suggest that there is an
interesting sense in which knowledge of other minds is direct. When people speak of direct
knowledge, they typically have observational knowledge in mind, but this need not be so.
For instance, if by ‘direct’ one simply means non-inferential it might also be argued that our
knowledge of logical and mathematical axioms is direct, or that our knowledge of our own
minds is direct. My concern here, however, will be with observational knowledge, since this
is McDowell’s main concern and it is the notion of directness that is relevant in the case of
other minds.5
My strategy is the following. I start, in sections 1 and 2, by trying my best to help
McDowell meet the challenge. I suggest a construal of McDowell’s account of direct or
observational knowledge, based on his discussion in Mind and World, which does have
epistemological significance and would, arguably, support the conclusion that knowledge of
other minds is direct. Indeed, I suggest that this is the only way of making McDowell’s claim
that knowledge of other minds is direct plausible. In section 3 I turn to a critical examination
of the suggested account of direct knowledge. I argue that further modifications are needed if
it is to qualify as an account of observational knowledge, and that once these are added it
becomes clear that there is an important difference between ordinary observational
knowledge and knowledge of the mental states of others. Moreover, I argue, it is a mistake to
assume that in order to reject the traditional inner-outer picture we must construe knowledge
of other minds in terms of direct observation. The conclusion is that although there is a sense
in which knowledge of other minds is direct, this falls short of McDowell’s claim that
knowledge of the mental states of others qualifies as observational.
1. Direct Knowledge
1.1 A question of phenomenology?
One common suggestion is that what characterizes observational judgments is that they have
a certain phenomenology -- they are, as it were, ‘cognitively spontaneous’. For example,
4
McDowell 1982, 456. See also McDowell 1978, 136: On certain occasions, McDowell argues, “one can
literally perceive, in another person’s facial expression or his behaviour, that he is in pain, and not just infer that
he is in pain from what one perceives.”
5
This is not to say that McDowell always uses the notion of direct knowledge as synonymous with
observational knowledge. For instance, McDowell argues that we can have direct knowledge of the past
(through memory) but he would of course not claim that we can have perceptual knowledge of the past
(McDowell 1978.)
3
when I judge “There’s a red book on the table” I do not reach this belief through an
argument, rather, the belief just occurs to me. Inferential knowledge, by contrast, involves
explicit argument from premises to conclusions.
But this cannot plausibly be all there is to the distinction between direct and indirect
knowledge. Construed this way the distinction lacks philosophical interest since it suggests
that whether a piece of knowledge is inferential or not is just a matter of individual
psychology. Thus, if a theory is sufficiently internalized anything can be observable in this
sense (for instance, the well-trained physicist will “perceive” sub-atomic particles), and there
is nothing surprising or interesting about the claim that we can have knowledge of other
minds through observation (most of the time I do not go through any conscious inferences
when I judge that someone else is in pain). To be of interest the distinction should concern
not how we in fact happen to reason, but how we would reason if our claims to knowledge
were challenged.
I mention this construal of the distinction since McDowell does stress
phenomenological considerations when arguing against inferential models, and this may give
the impression that he construes the distinction between direct and indirect knowledge
phenomenologically. For example, McDowell argues against the view that in learning a
language one puts a theoretical construction on what one ‘really’ perceives (facts about
linguistic behaviour) on the grounds that such a view would “falsify the phenomenology of
understanding speech in a familiar language.”6 That is, McDowell seems to suggest that an
epistemology of understanding which makes knowledge of what others people say inferential,
theory-dependent, is unacceptable since we do not go through any conscious inferences when
understanding speech. Similarly, in a discussion of Dennett’s account of perceptual
experience, McDowell criticizes Dennett for failing to capture the phenomenology of
perception, and argues, instead, for a model of perception according to which perceptual
experiences involve direct encounters with external objects.7
Despite remarks of this sort it is clear that when McDowell rejects inferential models
he is not merely making a point about phenomenology, but an epistemological one. Thus, in
his discussion of perceptual experience McDowell makes clear that a bad phenomenology is
an objection only if it necessitates a bad epistemology.8 And in his discussion of other minds,
McDowell argues: “[T]he rejection of the inferential model does not turn on mere
6
McDowell 1981, 238.
McDowell 1994. Reprinted in McDowell 1998, 341-358. For a critical discussion of McDowell’s emphasis on
phenomenological considerations see Bilgrami 1992.
7
4
phenomenology. Theory can partly ground a claim to knowledge even in cases where it is not
consciously brought to bear; as with a scientist who (as we naturally say) learns to see the
movements of particles in some apparatus.”9
If, therefore, the distinction between direct and indirect knowledge is an
epistemological distinction, how is it to be understood? To answer this question let us take a
closer look at Mind and World where the notion of an observational judgment stands at the
centre.10
1.2 Minimal Empiricism
In Mind and World McDowell spells out what he characterizes as a “minimal empiricism”.
The traditional empiricist account is familiar. This account rests on the notion of what is
given in experience, prior to any conceptualization or theorizing. Observational judgements
are said to be non-inferential in the sense that they are justified directly, by what is given in
experience, whereas non-observational judgments require for their justification reference to
theory as well. Like the traditional empiricist McDowell speaks of what is given in
experience and holds that what is thus given provides a direct justification for observational
judgments. However, he opposes the construal of the given as being prior to
conceptualizations. If the given is construed that way, McDowell argues, it falls outside the
“space of reasons” and cannot play a justificatory role. The solution is to construe what is
given in experience as something which is already conceptualized. Experiences, McDowell
says, “are impressions made by the world on our senses, products of receptivity; but those
impressions themselves already have conceptual content.”11 This allows experiences to play a
genuinely justificatory role and put us in direct touch with reality:
That things are thus and so is the content of experience, and it can also be the
content of a judgement: it becomes the content of a judgement if the subject
decides to take the experience at face value. So it is conceptual content. But
that things are thus and so is also, if one is not misled, an aspect of the layout
of the world: it is how things are. Thus the idea of conceptually structured
operations of receptivity puts us in a position to speak of experience as
8
McDowell 1998, 344.
McDowell 1982, 478, fn 1.
10
McDowell 1994. See especially chapters 1 and 2.
11
Ibid. 46.
9
5
openness to the layout of reality. Experience enables the layout of reality itself
to exert a rational influence on what a subject thinks.12
McDowell’s minimal empiricism therefore appears to provide us with a very straightforward answer to our question: A piece of knowledge is direct, observational, if it is justified
by experience alone, whereas a piece of knowledge is indirect if it requires for its justification
a set of beliefs as well. However, matters are a bit more complicated. The complication
derives from McDowell’s claim that experiences are conceptualized.
That experience is conceptualized, McDowell says, implies that experiences, although
passive, are caught up with spontaneity, with active thought. This means that experiences too
are susceptible to critical thinking, and that all judgments require a background
understanding, a network of beliefs. Consequently, “even the most immediately observational
concepts are partly constituted by their role in something that is indeed appropriately
conceived in terms of spontaneity.”13 But this puts pressure on McDowell’s account of
observational judgments. If all judgments are caught up with spontaneity, it would seem,
then no judgments are justified by “experience alone”. Even a colour judgment such as
“That’s red” requires a background of beliefs without which the judgment would not be
justified. Moreover, McDowell subscribes to the holistic nature of the conceptual sphere.
The capacities that are drawn on in experience, McDowell argues, “are recognizable as
conceptual only against the background of the fact that someone who has them is responsive
to rational relations, which link the contents of judgments of experience with other judgeable
contents. These linkages give the concepts their place as elements in possible views of the
world.”14 Thus, not only does an observational judgment require a set of background beliefs,
but this set is caught up with the larger network of beliefs that go into our view of the world.
If this is so, however, how can the sharp distinction between observational and nonobservational judgments be upheld?
McDowell hints at two possible replies to this question. The first consists in an appeal
to the distinction between ‘active’ and ‘passive’ thought. What characterizes observational
judgments, on this view, is their passivity, that they do not involve any active reasoning. 15
But this reply is not very promising. The claim that observational judgments are passive does
not help us to pick out an epistemically interesting category of judgments, any more than
does the suggestion that observational judgments have a certain phenomenology, since even a
12
Ibid. 26.
Ibid. 13
14
Ibid. 12.
13
6
highly theoretical judgment (such as the scientist’s) can appear quite passive (as when the
theory in question is internalized). Something further is therefore needed if McDowell is
going to be able to give the notion of direct or observational knowledge some epistemic
significance.
The second reply consists in appealing to a distinction between conceptual holism and
justificatory holism. To have the concept red, it could be argued, I need to have all sorts of
other concepts, other beliefs, but my judgment “This is red” is not justified by these other
beliefs but, rather, by the fact that I experience that the object in question is red.16 What
justifies my observational judgments is therefore not further beliefs, even though in order to
make such a judgment a host of other beliefs needs to be in place. The trouble is that it hard
to see how McDowell can appeal to a distinction of this sort. The conceptual realm, on
McDowell’s view, is the realm of ‘space of reasons’. A concept is to be understood in terms
of the role it plays in our judgments, in terms of inferences and rational interconnections
between beliefs. If the conceptual is characterized this way, it should be clear, conceptual
holism will imply justificatory holism. In fact, McDowell himself grants this. Responding to
the claim that his position is foundationalist, McDowell suggests that although he is a
foundationalist in the sense that he credits experience with a warranting role, the image of
foundations is potentially misleading:
There is a rational dependence of the ‘superstructure’ on the ‘foundations’.
But just because concepts are involved in experience, and the conceptual
realm is a seamless web of rational interconnections, there is also a rational
dependence (of a different sort) in the opposite direction. The ‘foundations’
are partly held in place by the ‘superstructure’.17
It is therefore unclear how McDowell can hold both that experience is conceptualized
and that there is a sharp distinction between observational and theoretical judgments. It is of
no help, notice, to respond by saying that observational judgments are those that report the
content of an experience. We also need to be assured that there are some limitations on what
can be experienced. Conceptualizing experience allows McDowell to expand the range of
what can be experienced beyond the range of the traditional empiricist observables, but the
question is whether there are any limits to this expansion. That is, once experience is
conceptualized, once it is made essentially propositional, is there anything which cannot be
15
16
Ibid. 10.
McDowell does seem to suggest something like this. See ibid. 12 and 30, and McDowell 1993.
7
experienced? For example, what is there to prevent us from saying that the scientist perceives
sub-atomic particles (he perceives “that the world is thus and so”) and that the judgment he
forms is observational?
The question, thus, is how McDowell can give the distinction between direct and
indirect knowledge some epistemic weight, without falling back on the notion of a nonconceptualized given. That is, assuming that experience is conceptualized, and that concepts
are to be understood holistically, how can we avoid the conclusion that the distinction
between observational and theoretical knowledge dissolves? How are we to preserve the idea
of a ‘foundation’ and a ‘superstructure’?
1.3 Minimal Empiricism Modified
The most promising strategy if one is to remain faithful to McDowell’s framework, I believe,
is the following: The mistake above, it can be argued, is to assume a form of ‘rampant’
conceptual holism. What we should do, instead, is limit the holism, and we can do this by
claiming that the theory presupposed by observational reports is wholly unlike scientific
theories, ‘theories proper’, and enjoys a special status in our network of beliefs. It is then
possible to distinguish the observational from the non-observational by saying that while
observational knowledge is theory-dependent it is only dependent on theory in the basic, lowprofile sense. For instance, to justify my judgment “There’s a red book on the table” a
background of common sense beliefs needs to be in place, but to justify the judgment “There
are electrons in the apparatus” a full-fledged, disputable, scientific theory is required. In this
sense “There is a red book...” is observational whereas “There are electrons...” is not. Thus,
what characterizes observational judgments is that their theory-dependence is limited to the
very basic ‘theory’, the common-sense conception of the world, and the danger of
undermining the distinction between the observational and the non-observational has been
averted.18
Now, I think that there is something appealing about this strategy. Distinguishing
between ‘basic’ theories and ‘proper’ theories allows us to grant that all observation is theory
dependent, without having to completely give up on the idea of relatively theory-independent
observational data that can be used to test and compare theories proper: That is, on this view,
17
McDowell 1996, 284. See also McDowell 1993, where McDowell grants that the label ‘coherentism’ fits his
view better than ‘foundationalism’.
18
The idea, obviously, cannot be that everything that is in the basic theory is observational (only a fraction of
our common sense view of the world could plausibly be said to be observational). Rather, the idea would have
8
all observations are dependent on basic theory, our common sense view of the world, but
since the basic theory is distinct from theories proper the observational data have a relative
independence from the latter type of theory.19 Of course, the notion of ‘basic theory’ is quite
vague and the suggestion need not be that there is a sharp distinction between basic and nonbasic theories. It can be granted that there is an interdependence between ‘foundation’ and
‘superstructure’, such that the theory operative at the foundational level may, on occasion, be
revised in the light of the explanatory theories of the superstructure. Nonetheless, this has to
be the exception or else the very idea of observational evidence is obliterated. Thus, although
the notion of a basic theory is vague, what characterizes a basic theory is that it constitutes
the unquestioned background against which theories proper are formulated and tested. A
large-scale revision of the basic theory would therefore be very unlike a large-scale revision
of theories proper -- it would imply a revision not only in our explanations of observed
phenomena, but in our observational judgments, the ‘evidence’.
Although McDowell does not talk explicitly in these terms, there are reasons to
believe that a view of this sort should be congenial to him. First, it would allow him to grant
that experience is conceptualized and observational judgments theory-dependent while, at the
same time, making sure that the notion of observational knowledge has some epistemological
significance. Second, at points McDowell does seem to suggest that observational judgments
depend on something like a basic theory. Consider his discussion of conceptual schemes. 20
Conceptual schemes, McDowell argues, are not to be understood as a product of reason as
opposed to nature, but, rather, as part of our ‘second nature’. By being acculturated, initiated
into a ‘tradition’, human beings naturally acquire certain conceptual capacities, a
responsiveness to reason and this is simultaneously an initiation into the world. Conceptual
schemes are therefore not a theory about the world but, rather, “constitutive of our
unproblematic openness to the world.”21 Once initiated into the sphere of reason one can
extend one’s cognitive grasp by formulating theories proper about the world encountered, but
to be that what characterizes observational judgments is that, although they are theory-dependent, their theorydependence is limited to the ‘basic theory’. I return to this issue below in section 3.
19
Bruce Aune suggests an account of this sort. He argues that when we test theoretical claims by observational
data “we are actually relating theory to theory” (1986, 565). See also Guttenplan 1994, 22.
20
McDowell 1994, Lecture IV and pp. 155-159.
21
McDowell 1994, 155. This idea can be found already in McDowell 1981: “in acquiring one’s first language
one acquires a conception of the world” (242).
9
such theories presuppose that something much more basic, more ‘natural’, is in place: the
conceptual scheme or ‘world view’ which makes the world available to us in the first place.22
It is also clear that something like this is needed if McDowell’s claim that knowledge
of other minds is observational is to be made at all plausible. After all, our attributions of
mental states to others is dependent on a set of background beliefs about the interconnections
between mental states, expressions, and actions; i.e. the set of background beliefs commonly
referred to as ‘common sense psychology’. For instance, when we judge “She is in pain” of a
person who is injured and screams out, assumptions about the causes and expressions of pain
are essential. The claim that knowledge of other minds is observational, therefore, better not
be understood as implying a denial of the idea that common sense psychology plays this
crucial role.23 How, then, is it to be understood? Equipped with the distinction between basic
theories and theories proper we have an answer: Knowledge of other minds is observational if
common sense psychology is not a theory proper but plays the role of a basic theory, rather
like our common sense view of the world as existing of macroscopic, enduring objects plays
the role of a basic theory.
Let us therefore assume, provisionally, that this provides us with a notion of
observational knowledge which has epistemic significance and is compatible with
McDowell’s framework. The question whether we can have direct knowledge of other minds
has then been converted into a question concerning the status of common sense psychology.
Our judgments about the mental states of others depend on background ‘theory’, on common
sense psychology, but what is the status of this theory? Is it akin to the explanatory theories
of the sciences, a theory ‘proper’, or is it more like our common sense view of the world?
When McDowell criticizes the claim that knowledge of other minds is theoretical this is also
how he construes the issue. Discussing acquisition of mental terms he says:
Acquiring mastery of the relevant tracts of language is not, as acquiring a
theory can be, learning to extend one’s cognitive reach beyond some previous
limits by traversing pathways in a newly mastered region of the ‘space of
reasons’. It is better conceived as part of being initiated into the ‘space of
reasons’.24
22
Of course, McDowell would probably not like to use the term ‘theory’ in this context. (See McDowell 1981,
239). However, since ‘theory-dependence’ has become the standard term when expressing the idea that all
judgments depend on a background of beliefs, I shall continue to speak of a ‘basic theory’.
23
Consider McDowell’s claim that we can have direct perception of the content of what other members of the
community say (1981). Quite clearly, McDowell would grant that this ‘direct perception’ presupposes a huge set
of background assumptions, both about the other’s language and psychology.
24
McDowell 1982, 477.
10
That is, according to McDowell, mental terms cannot be said to be theoretical since acquiring
these terms is part and parcel of being initiated into the space of reasons, and not part of
theorizing proper where one’s cognitive grasp is extended. How compelling is this
suggestion?
2. Inferring Other Minds
2.1 McDowell on Criteria
McDowell’s opposition to inferential models in the case of other minds, is most explicit in his
critical discussions of the so called ‘criterial view’.25 According to the criterial view, recall,
the relationship between inner and outer is inferential, but the inference is of a special, noninductive kind. This is so, it is argued, since in the case of other minds the relation between
evidence, i.e. behavioural criteria, and mental state is conceptual or conventional. This does
not, however, prevent the relation from being defeasible, as when P merely pretends to be in
pain. Thus, it is suggested, there is a defeasible, conceptual connection between inner and
outer.26 McDowell rejects the criterial view on the grounds that it is incoherent. The trouble,
McDowell argues, derives from the defeasibility of criteria. If all we ever experience is the
satisfaction of criteria, and if experiencing this is compatible with the falsity of “He is in
pain” then we are experiencing a fact which falls short of experiencing that the other is in
pain: “The fact itself is outside the reach of experience.”27 How then, McDowell asks, can
experiencing the satisfaction of criteria constitute knowing that the other is in pain? He
concludes: “That yields this thesis: knowing that someone else is in some ‘inner’ state can be
constituted by being in a position in which, for all one knows, the person may not be in that
‘inner’ state. And that seems straightforwardly incoherent.”28 To resolve the incoherence,
McDowell suggests, we should give up the assumption that criteria are defeasible and say,
instead, that the circumstance that someone else is in a certain mental state can be directly
observed, without any epistemic intermediaries.
What does the alleged incoherence consist in? It is of course incoherent to claim that
one knows that P is in pain and yet it is not true that P is in pain. However, McDowell could
hardly be charging the criterial view with such a blatant mistake. That criteria are defeasible
does not mean that one can know that P is in pain when he is not, but merely that one can be
justified in concluding that P is in pain (because criteria are fulfilled) when he is not. Where,
25
McDowell 1978 and McDowell 1982.
See for example Lycan 1971.
27
McDowell 1982, 457.
26
11
then, is the incoherence?
McDowell’s suggestion seems to be the following: The
defeasibility of criteria implies that M itself, the mental state, falls outside the range of my
experience, which would suggest that knowledge of other minds is inferential, based on
theory. However, the connection between criteria and M is not supposed to be inferential in
the ordinary sense but conceptual. But if the connection is conceptual then how can it fail to
hold? What is incoherent, in other words, is not the suggestion that criteria are defeasible per
se, but this suggestion in combination with the claim that there is a conceptual connection
between criteria for M and M itself.29
Put this way, McDowell’s objection does point to a serious difficulty with the criterial
view, since it is not clear what we are to make of the notion of a defeasible conceptual
connection. It is of course true that there is a conceptual connection between pain-behaviour
and pain. But this connection is not defeasible. And it is true that someone can scream,
grimace, etc., without being in pain, but there is no conceptual connection between grimacing
and pain. What the criterial view ignores, it might be said, is the point emphasized by
Davidson, that conceptual connections hold between descriptions.30 Whether the connection
between X and Y is conceptual or not depends on how X and Y are described. Described one
way there is a conceptual relation (sunburn - sun), described another there is not (red skin sun). But there is no such thing as a defeasible conceptual relation.
McDowell is therefore quite right to question the criterial view. The question is why
we should resolve the incoherence along the lines recommended by McDowell, by making
knowledge of other minds observational. Another option is simply to reject the assumption
that there is a conceptual connection between inner and outer,31 and construe knowledge of
other minds as straightforwardly inferential, along the lines of theoretical knowledge in the
sciences. It is time to consider this option in some detail.
2.2 An Inference to Best Explanation?
According to the prevailing view today knowledge of other minds is a prime example of
inferential knowledge. The mental states of others are unobservable, it is held, but this does
not make knowledge of other minds any more problematic than knowledge of unobservables
28
Ibid., 457.
See ibid. 459: “To hold that theory constitutes to the epistemic standing, with respect to a claim. . . . would
conflict with the insistence that ‘criteria’ and claim are related by ‘grammar’; it would obliterate the distinction
between ‘criteria’ and symptoms.”
30
Davidson 1963.
31
This reaction to McDowell’s criticism of the criterial view is recommended by Bilgrami 1992, 322.
29
12
in the sciences, such as our knowledge of subatomic elements in physics. 32 In both cases we
rely on an inference to best explanation: Just as the postulation of micro-elements is justified
insofar as it explains the behaviour of macro-elements, the postulation of mental states and
events is justified if it explains the behaviour of others. Mental terms, it is therefore
suggested, are theoretical terms much like ‘electron’ or ‘lepton’ -- they are given their
meaning by the role they play in common sense psychology, a psychological theory which
serves to explain and predict behaviour. The traditional worry that the unobservability of
other minds must lead to epistemic and semantic difficulties, it is argued, is simply based on a
primitive view of the role of unobservables in the sciences. 33
Since, according to this view, the connection between inner and outer is straightforwardly inductive McDowell’s criticisms of the criterial view do not apply. Why then does
McDowell insist on the observational model? The answer had better not be that theoretical
terms in general are semantically suspect. That would amount to a form of verficationism.
Rather, the claim must be that there are special reasons why common sense psychology
cannot be said to be a theory proper and mental terms cannot be compared to the theoretical
terms of science. What might these reasons be? Although McDowell does not give a very
clear answer,34 let us consider two such reasons.
The first has to do with the fact that mental concepts have a first person use. In order
for the suggestion that mental concepts are theoretical to be plausible, it must be claimed that
this holds both of the first and third person use. Otherwise, one would fail to account for the
unity of mental concepts; for the idea that ‘pain’ expresses the same concept when applied in
the first and in the third person case, and unless this unity can be secured, quite clearly, no
progress can be made with the epistemological problem either (if ‘pain’ cannot express the
same concept in the first- and third person case there could be no such thing as knowing that
the other is in pain). Consequently, the suggestion that mental terms are theoretical must
apply not only to the third person but to the first person use as well.35 But this is very
32
For a classic statement of this view see Sellars 1956. For more recent versions see Aune 1986, Churchland
1984, Ginet 1985, Pargetter 1984, Stemmer 1987.
33
Churchland: “The problem of other minds was first formulated when our grasp of the nature of theoretical
justification was still rather primitive.” (1984, 70)
34
At one point McDowell suggests that the real objection to the ‘theory-theory’’ approach concerns concept
acquisition: If all we ever experience in the case of others were mere behaviour, McDowell says, then we could
never acquire the mental concepts and so could never be in a position to formulate a theory about others (1982,
477). The problem with this suggestion is that it relies on assumptions about concept acquisition that are
difficult to assess. Why should it be any more difficult to acquire the mental concepts on the basic of behaviour
than, say, the concept of a molecule on the basis of behaviour of physical substances?
35
This is typically granted by those who defend the idea that mental concepts are theoretical. See Aune 1986,
566, Churchland 1984, 76-79, and Sellars 1956, 195.
13
implausible. It is implausible not because first person judgments do not seem to involve any
theory (that would be a mere phenomenological point), but because of the fundamental role
played by first person judgments. What characterizes theoretical terms is their principled
dispensability; if a better explanatory theory is made available then we can dispense with the
term and, by implication, the concept. But how could we abandon the first person use of
mental concepts? Doing so would involve not just giving up the belief in a certain kind of
explanation, but giving up all the judgments that make up the first person point of view (I am
in pain, I am hungry, I think it is going to rain, etc. etc.). That is, abandoning the first person
use of mental concepts would involve reconstituting our inner lives, and this suggests that
mental concepts are utterly unlike theoretical concepts of the sciences. 36 Acquiring these
concepts is part and parcel of coming to be a human being with an inner life and is therefore
not to be conceived of as “extending one’s cognitive reach”, to use McDowell’s phrase, but
rather as part of “being initiated into the sphere of reasons”.
Of course, there are those who are happy to grant that we can dispense with the first
person use of mental concepts, i.e. eliminativists such as Paul Churchland. 37 This is not the
place to address the debate over eliminativism. What can be said is that the comparisons
drawn by eliminativists between common sense psychology and earlier, discarded theories
such as alchemy are very strained. Abandoning alchemy meant abandoning a theory which,
although influential among scientists of its day, played no role in our ordinary lives an
practices. Abandoning common sense psychology, by contrast, would imply a reconstitution
of our ordinary practices and lives on a scale never encountered before.38
This takes us to the second reason one might suspect common sense psychology is not
a theory proper. It was suggested above that a criterion of theories proper is that observational
data are relatively independent of the theory in question. This is true in the case of alchemy
and modern chemistry, where the theories could be construed as competing accounts of a
shared set of observational data. However, crucially, common sense psychology does not
meet this criterion. If one were to abandon common sense psychology altogether, and with it
all intentional notions, the ‘explanandum’ of common sense psychology would have to be
reconstituted as well -- what would be explained would no longer be human actions but
bodily motions. Thus, there could be no evidence in common between common sense
36
Sellars, in fact, acknowledges this. Although mental concepts are tied up with the theoretical/observational
distinction, he argues, “it would be paradoxical and, indeed incorrect, to say that these concepts are theoretical
concepts.” Sellars 1956, 183.
37
For instance, Churchland suggests that “even our introspection may be reconstituted within the conceptual
framework of completed neuroscience. . ..” (1981, 67).
14
psychology and a theory which eliminates all references to the psychological and this too
suggests that common sense psychology plays the role of a basic theory.39 The point might be
put as follows: Knowledge of other minds cannot be said to be an inference to best
explanation, since that which is to be explained (human actions) already presupposes that
there are other minds. In this sense there is an important disanalogy between the case of other
minds and the case of theoretical entities in the sciences.
Hilary Putnam makes this point in a discussion of other minds.40 In one sense, Putnam
suggests, our acceptance of the proposition that other people have mental states is analogous
to the acceptance of ordinary empirical theories -- part of the justification of the proposition
is that to give it up would require giving up all of the statements that imply that proposition
(“He is in pain”, “She thinks that it is raining”, etc. etc.). However, Putnam continues, there is
also an important disanalogy with ordinary empirical theories: It is built into the language
used to make observation reports (“He screamed”, “She took her umbrella”, etc.) that other
people have mental states. Our reasons for accepting that others have mental states, Putnam
argues, are therefore not an ordinary induction any more than our reasons for accepting that
material objects exist are an ordinary induction. That is, neither the belief that there are
external objects, nor the belief that there are other minds can be said to be based on an
‘inference to best explanation’, since that which is explained by the ‘theory’ (the behaviour of
macroscopic objects and the actions of human beings, respectively) already presuppose the
truth of the beliefs in question.
There are therefore reasons to be skeptical of the claim that common sense
psychology plays the role of a theory proper, and that knowledge of other minds is based on
an inference to best explanation. If this is right we have found support for McDowell’s claim
that knowledge of other minds is, in a sense, direct: Although our judgments about others
involve a ‘theory’, this theory does not play the role of a scientific theory but a much more
fundamental one, akin to our common sense view of the world.41
3. Observing Other Minds
38
See Horgan and Woodward 1985 for a discussion of this.
For this reason it is odd to argue, as the eliminativist does, that folk psychology should be abandoned since
there are (will be) competing neurophysiological theories that explain the evidence better. See Churchland 1981
and Stemmer 1985.
40
Putnam 1975.
41
This, I believe, is related to Wittgenstein’s emphasis on the role of natural expressions: The idea that pains
have certain expressions is not part of a speculative theory (nor is it a ‘convention’) but something much more
primitive; it belongs with the framework that makes our language possible.
39
15
My strategy has been to construe McDowell’s notion of observational knowledge as
knowledge involving basic theories only, and then to argue that common sense psychology is
a basic theory. This, I believe, is the only strategy available if one is to make it at all plausible
that knowledge of other minds is direct, given the obvious dependence of judgments about
the mental states of others on common sense psychology. The question now is whether this
really shows that knowledge of the mental states of others is observational. That is, even if
we grant that knowledge of other minds is direct in the sense suggested above, does this
notion of ‘directness’ really capture all there is to observational knowledge?
It would seem not. Consider the case where a person pours a glass of water and
drinks it, and we judge “She is thirsty”. Would we really want to say, in this situation, that we
observe that the other person is thirsty? The more intuitive claim, it would seem, is that we
observe that she drinks a glass of water and that, on the basis of this observation, we infer that
she is thirsty. Similarly, in the case of a person who touches a hot plate and pulls away her
hand. What we observe, it would seem, is what she did, not what she feels, even though
nothing but a basic theory is involved.
We need to make some further modifications of the account of observationality
suggested above. Let us grant that a necessary condition on observational judgments is that
justification is limited to basic theory. This, again, is plausible since it allows us to say that all
observation is theory dependent without undermining the distinction between the
observational and the theoretical altogether. However dependence on basic theory cannot
plausibly be a sufficient condition, since even among the judgments whose justification is
limited to basic theory some are clearly less observational than others. As often pointed out,
observational judgments have a certain ‘last ditch’ quality, it’s what we fall back on in the
case of disagreement, and not all judgments dependent on basic theory have this quality.
Thus, we may disagree on whether the other is thirsty without disagreeing on what she did,
and we are much more likely to revise the former belief than the latter. In this sense the
judgment “She is thirsty” is less observational than the judgment “She poured a glass of
water and drank it”, and “She screamed” is more observational than “She is in pain”.42
If this is right we should not simply divide our empirical judgments into two
categories, the observational and the theoretical ones. Rather, our judgments fall into three
categories: Observational judgments (judgments whose justification is limited to the basic
42
This point is made by Blackburn 1992, 193.
16
theory and have a last-ditch character),43 quasi-theoretical judgments (judgments whose
justification is limited to the basic theory but are not last-ditch), and theoretical judgments
proper (judgments whose justification require reference to theories proper). It is then clear
that not all judgments about other minds can be treated in the same way. Thus, it is
reasonable to suggest that what typically goes into explanans of common sense psychology is
less observational than explanandum. In short, what people do is more observable than what
they feel and experience.44 Of course, because of the holism there are no sharp separations to
be made here. Sometimes revising explanans will lead us to revise explanandum (she did not
scream, she was merely practising her scales), and there are situations in which we are as
unwilling to revise explanans as revising explanandum (as when a person has just been
injured and screams out). Nonetheless, we should grant that some third person judgments are
less observational than others.
In particular, I think we should be skeptical of McDowell’s suggestion that
knowledge of what others say when using language is observational.45 Although
phenomenologically our knowledge of what others are saying is often ‘direct’, such
judgments cannot plausibly be said to have the last-ditch character of observational
judgments.46 Rather, when interpreting another speaker we try to make sense of her
utterances (and actions), and when this fails, the judgments we are most likely to revise are
judgments about the content of what she said. We conclude, for instance, that the other
cannot have meant what she said (she made a slip of the tongue) or that her words must mean
something different from what we assumed (she uses her words non-standardly). What plays
the fundamental role when trying to make sense of the other, therefore, is not judgments
about the contents of what she is saying, but judgments about her actions, such as what
sentences she holds true and false, what she infers from what, etc. These actions, as
emphasized by Davidson, lie at the basis of inter-personal communication, they provide the
entry-points into the other’s language, and must be considered more observational than
judgments about the content of what other people say.47
43
This may still not give us a sufficient condition on observationality since there seem to be judgments of this
sort which are not intuitively observational (for example, “I am a human being”), but I will not pursue the issue
further here.
44
I am grateful to Kathrin Glueer for valuable discussions of this issue. Samuel Guttenplan takes a similar line:
“It is easier to see what people do than what they believe, but it is also easier to see what they believe than what
they experience.” (1994, 17).
45
See especially McDowell 1981, 241-242.
46
This point is made by Blackburn 1992, 193.
47
Davidson 1973. Of course, we may later come to revise our construal of the evidence, as a result of our
interpretational theory (see Bilgrami 1994, 218, for a discussion of this). For instance, we conclude that what
we took to be a holding true was in fact an insincere assertion. The point is simply that this must be the
17
There is therefore an important difference between ordinary observational knowledge,
such as knowledge that there is a red book in front of me, and knowledge of the mental states
of others. The difference turns on the fact that the latter type of knowledge involves trying to
make sense of what the other does and says. This, again, is not to say that knowledge of other
minds is an inference to best explanation, in the sense that knowledge of theoretical entities in
the sciences are, since that which is to be explained (human actions) already presupposes that
the other has a mind. That is, I do not believe that the other has a mind because this best
explains my observable data (bodily motions); rather, I assume that she has a mind, that what
I am observing is human actions, and go on to try to make sense of her actions. If I fail to
make sense of the other, I may question my initial assumption that what I am encountering is
a human being, but this does not mean that my assumption that she is a human being is based
on an inference from observation of bodily motions. It is therefore possible to deny that
knowledge of other minds is straight-forwardly observational without making any
concessions to the traditional idea that all that is ever available from the third person point of
view is mere bodily motions.
However, McDowell resists this line of reasoning and argues that we should treat
knowledge of all mental phenomena in the same way. Consider his discussion of how one can
know that someone else has a red image. The answer “By what he says and does”, McDowell
argues, should not be construed as appealing to a condition that is more certain than the
judgment “He has a red image” itself : “a condition that someone might satisfy even though
he has no red image, so that it constitutes at best defeasible evidence that he has one.”48
McDowell’s worry is that if we were to construe the evidence this way, as being more
accessible and certain than that which it is evidence for, we would be stuck with the
traditional inner-outer picture and the skeptical predicament. But, again, this worry is
ungrounded. To say that we may be more certain of what a person does than of what he
experiences is not to say that all we ever have access to is bodily motions, mere behaviour,
and there is no implication that the minds of others are forever beyond our reach.
To illustrate this further consider McDowell’s discussion of how we are to account for
mistakes, as for example when we mistake pretended pain for real pain. McDowell argues
that it is crucial to reject the assumption that there is something in common between the
veridical case and the deceptive case. To say that there is something in common, he suggests,
exception, and that in general we must be able to discern these basic attitudes before we have developed a
theory of interpretation.
48
McDowell 1982, 465.
18
is to say that even in the non-deceptive case we experience something which “falls short of
the fact ascertained”49; it would follow that all that is ever available to experience is bodily
motions, ‘psychologically neutral information’, and we would be back with the inner-outer
picture and the skeptical problem. In order to avoid this, McDowell argues, we must endorse
a disjunctive conception according to which an experience that P is in pain is either a mere
appearance or the fact that P is in pain making itself manifest. In the veridical case, thus, it is
not that we draw an inference from what is directly experienced (observable behaviour) to
what is not (the inner state); rather, we have direct experience of the fact that the other is in a
certain inner state.
But this move, too, is unmotivated. It is quite possible to say that there is something in
common between the veridical and the non-veridical case, without falling back on the
traditional inner-outer picture: What is common between the two cases is not a bodily motion
but that the other acts as if in pain. This is not psychologically neutral information, in
McDowell’s sense (it employs both the concept of an action and the concept of pain) but it is
nonetheless something which ‘falls short of the fact’, i.e. of P being in pain. We can then say
that there is something in common between the two cases, without having to worry that this
implies that all that is ever available to observation is bodily motions.50
This is not to dispute the importance of a ‘disjunctive epistemology’ altogether.
McDowell has also applied the disjunctive model in the case of perceptual illusions, and it
may well be that he is right to suggest that in that case we need to appeal to a disjunctive
model to avoid the implication that all we ever observe is appearances, sense-data.51 But it is
clear that in the case of other minds we need not appeal to the disjunctive model in order to
avoid the implication that all we ever observe is bodily motions, since what is claimed to be
in common between the veridical and the non-veridical case is human actions. We should
therefore question the comparison with perceptual illusions. Making a mistake about the
mental states of others is not like an illusion, but more like a misinterpretation: It is not a case
of seeing something which is not there, but a case of misunderstanding that which one sees is
there (the action). It may be that there is a kind of mistake in the case of other minds which is
49
Ibid., 472. See also ibid. 466-467.
McDowell makes one comment that might give the impression that he would grant that some mental states are
less observational than others. He suggests that in the case of some mental states (emotions, for example), the
model of direct observation works better than in the case of other mental states (such as pains). In the latter case,
he suggests, we should say that what is directly available to experience is not the inner state itself, but the person
“giving expression to his being in that ‘inner’ state” (1982, 473). However, McDowell makes quite clear that
one cannot give expression to being in such a state unless one actually is in such a state (since then what is
observable would “fall short of the fact”) and this means that he would resist the line suggested here.
51
McDowell 1986, 151.
50
19
similar to an illusion -- that of mistaking a robot for a human being. But it would be a serious
error to assimilate that kind of mistake to the types of ordinary mistakes we make when we
misunderstand other people. Indeed, that is precisely an assimilation that the skeptic tries to
force on us, and it should be resisted.
What can be concluded, then, is that McDowell is partly vindicated. There are reasons
to be skeptical of the claim that knowledge of other minds is based on an inference to best
explanation, along the lines of theoretical knowledge in the sciences, and so there is a sense
in which knowledge of other minds is direct. However, this falls short of McDowell’s
suggestion that knowledge of the mental states of others is observational. Such knowledge
involves making sense of what other people do and say, and in this respect it differs from
straight-forward observational knowledge. This is not something to be feared, but rather to be
embraced: Knowledge of the mental states of others is less observational and more fallible
than knowledge of objects such as tables and chairs, not because the mental states of others
are forever hidden behind bodily motions, but because human beings are infinitely more
complex in their capacities than ordinary observable objects.52
References
Aune, Bruce. 1986. “Other Minds after Twenty Years.” Midwest Studies in Philosophy, X.
Bilgrami Akeel. 1992. “Other Minds”. In A Companion to Epistemology, eds. J. Dancy and
E. Sosa. Oxford: Blackwell.
---- . 1994. “Dummett, Realism and Other Minds”. In The Philosophy of Michael Dummett,
eds. B. McGuinness and G. Oliveri. Netherlands: Kluwer Academic Publishers.
Churchland, Paul. 1981. “Eliminative Materialism and the Propositional Attitudes.” Journal
of Philosophy, vol. LXXVIII.
----. 1984. Matter and Consciousness. Cambridge, Mass.: MIT Press.
Davidson, Donald. 1963. “Actions, Reasons, and Causes.” In Essays on Actions and Events.
Oxford: Clarendon Press.
----. 1973. “Radical Interpretation”. In Inquiries Into Truth and Interpretation
Oxford: Clarendon Press.
Guttenplan, Samuel. 1994. “Stage 1: Mapping out the Territory.” In A Companion to the
Philosophy of Mind , ed. S. Guttenplan. Oxford: Blackwell.
52
Thanks to Kathrin Glueer and Barry Smith for valuable comments on an earlier version of this paper.
20
Horgan Terence and Woodward James. 1985. “Folk Psychology Is Here To Stay”.
Philosophical Review, XCIV.
Lycan, W.G. 1971. “Noninductive Evidence: Recent Work on Wittgenstein’s Criteria.”
American Philosophical Quarterly, vol. 8.
McDowell, John. 1978. “On ‘The Reality of the Past’”. In Action and Interpretation, eds. P.
Pettit and C. Hookway. Cambridge: Cambridge University Press.
----. 1981. “Anti-Realism and the Epistemology of Understanding”. In Meaning and
Understanding, eds. H. Parret and J. Bouveresse. New York, Berlin: W. de Gruyter.
----. 1982. “Criteria Defeasibility, and Knowledge”. Proceedings of the British Academy,
LXVIII.
----. 1985. “Functionalism and Anomalous Monism”. In Actions and Events: Perspectives on
the Philosophy of Donald Davidson, eds. E. LePore and B. McLaughlin. Oxford:
Blackwell.
----. 1989. “One Strand in the Private Language Argument”. 1989. Grazer Philosophische
Studien, 33/34.
----. 1993. “Knowledge by Hearsay”. In Knowing from Words, ed. B.K. Matilal and A.
Chakrabarti. Dordrecht: Kluwer.
----. 1994. Mind and World. Cambridge, Mass: Harvard University Press.
----. 1995. “Knowledge and the Internal”. Philosophy and Phenomenological Research, vol.
LV.
----. 1996. “Reply to Gibson, Byrne, and Brandom”. Philosophical Issues, 7.
Pargetter, R. 1984. “The Scientific Inference to Other Minds”. Australasian Journal of
Philosophy, 62.
Peacocke, Christopher. 1984. “Consciousness and Other Minds”. The Aristotelian Society,
supplementary vol. LVIII.
Robinson, Paul. 1991. “McDowell Against Criterial Knowledge”. Ratio IV.
Sellars, Wilfrid. 1956. “Empiricism and the Philosophy of Mind.” Minnesota Studies in the
Philosophy of Science, vol. I, eds. H. Feigl and M. Scriven.
Stemmer, Nathan. 1987. “The Hypothesis of Other Minds: Is it the Best Explanation?”
Philosophical Studies, 51.
21